Morphological differences are investigated using several culture strains of three sibling species collected from Taiwan and Guangdong in China and Pyinoolwin and Yangon in Myanmar. Careful examination of male terminalia reveals distinguishable differences in the paramere and the aedeagal basal process among the three species. In addition, a number of quantitative characters are compared. Kruskal-Wallis tests with Bonferroni correction, which are carried out separately for each sex, detect significant differences in 15 characters, of which two are male-specific, among the three species. Canonical discriminant analysis using these characters reveals that the three species can be distinguished from each other with high confidence for both sexes. The results clearly show the presence of three good species, Drosophila (Sophophora) lini Bock & Wheeler, 1972 and its two new siblings. The new species are described as Drosophila (Sophophora) ohnishii sp. nov. from Pyinoolwin and Drosophila (Sophophora) ogumai sp. nov. from Yangon. The morphological differentiation among the three sibling species does not coincide with the degree of reproductive isolation (based on a previous study). The premating isolation pattern suggests two possibilities that premating isolation has been evolved or reinforced in sympatric populations between D. ohnishii and D. lini and between D. ohnishii and D. ogumai or that it has evolved in a very restricted local population of D. ohnishii, possibly by a few mutations.
The Drosophila (Sophophora) montium species-subgroup is the largest in the D. melanogaster species-group, comprising a total of 85 species from Asia and Africa (Lemeunier et al., 1986; Toda, unpublished data). This subgroup includes a variety of species at different stages of speciation process and provides very useful materials for studies of evolutionary genetics. The relationships among members of this subgroup have been investigated by various methods; hybridization (cross experiment) tests (David et al., 1978; Kim et al., 1989), two dimensional electrophoresis and conventional starch gel electrophoresis (Ohnishi et al., 1983a, b; Ohnishi and Watanabe, 1984), chromosomal (metaphase karyotype) analysis (Baimai and Chumchong, 1980; Baimai et al., 1986), and eco-physiological comparison (Kimura, 1987). Through these studies several groups of very closely related species have been recognized as species-complexes: the D. auraria, the D. bakoue, the D. bocqueti, the D. jambulina, the D. kikkawai, the D. nikananu and the D. serrata complex. The members of each species-complex are hardly distinguishable in morphology and show various incipient stages in differentiation of characters such as genes, proteins, physiology, behavior and morphology. Those are very good materials for studying the speciation mechanisms.
A few genetic investigators have suggested the presence of two new species very closely related to Drosophila (Sophophora) lini Bock & Wheeler, 1972. Drosophila lini was described as a new species of the montium subgroup on the basis of a strain (Texas 3146.1) established from a single female collected from Taiwan in 1968 (Bock and Wheeler, 1972). When Tsacas and David (1977) established the kikkawai complex, they included D. lini in this species-complex.
Ohnishi and Watanabe (1984) was the first to recognize the presence of a sibling species of D. lini based on the results of two electrophoretic analyses. They analyzed protein difference by two-dimensional electrophoresis and allozyme variation by starch gel electrophoresis, using a strain (MMY326) originating from a single female collected at Pyinoolwin, central Myanmar in 1981, along with such strains of other 28 species of the montium subgroup. They placed MMY326 as a member of the kikkawai complex and tentatively named it as Drosophila lini-like. Then, Kim et al. (1989) examined the crossability between D. lini and D. lini-like (MMY326), and revealed that the premating isolation between them was almost complete. Furthermore, Kim et al. (1993) examined restriction enzyme patterns of mitochondrial DNA in 18 species of the montium subgroup, and placed D. lini-like (MMY326) closest to D. lini. More recently, Oguma et al. (1995) examined the reproductive isolation and courtship behavior between nine isofemale strains from Taiwan (Bowling Green stock no. 14028-0581.0=Texas old stock no. 3146.1=D. lini), mainland China (DHS315, DHS401, DHS501, NKS9231; all from Guangdong Province) and Myanmar (MMY326=D. lini-like, MMY307 from Pyinoolwin; RGN3, RGN206 from Yangon). They suggested the existence of at least three genetically distinct sibling species in them: D. lini (BG14028-0581.0, DHS315, DHS401, DHS501 and NKS9231) distributed from Taiwan to southern China, D. lini-like (MMY326 and MMY307) in central Myanmar and the other species (RGN3 and RGN206) in southern Myanmar. However, those species can be hardly distinguished from each other morphologically, and the two new sibling species have not been formally described yet.
In this study we compare the morphology precisely for both qualitative and quantitative characters among the three sibling species, using nine isofemale strains of which eight are the same as in Oguma et al. (1995), and describe them, two of them as new species.
MATERIALS AND METHODS
Ten male and 10 female specimens were examined for each of the nine isofemale strains (Table 1). The strains have been maintained on a standard Drosophila culture medium for long periods since the establishment each from a single wild-caught female.
Isofemale strains investigated
External morphology was observed under a stereoscopic microscope and metric characters were measured with an ocular micrometer. The detailed structure of head, male foreleg and male and female terminalia were observed in a droplet of glycerol under a compound light microscope, after detaching the organs from the body and cleaning them by warming in 10% KOH solution at about 100°C for several minutes. Drawings were made on the basis of microscope photographs taken by a computer-interfaced digital camera.
All type specimens were deposited in Systematic Entomology, The Hokkaido University Museum, Hokkaido University, Sapporo, Japan (SEHU).
Terminology and quantitative characters
We followed McAlpine (1981) for morphological terminology and Zhang and Toda (1992) for the definitions of measurements and indices. A total of 31 quantitative characters were measured and/or calculated for both sexes, 10 characters for male, and one character for female (Table 2).
Definitions of quantitative characters examined
First, to find meaningful characters for discriminating the three sibling species, we intensively examined three strains (BG14028-0581.0, MMY326 and RGN206), one from each species, for the above 42 quantitative characters. As some metric characters would have been affected by the body size, i.e. culture conditions, correlation with the thorax length (ThL) was examined for all characters except for the body length, the wing length and the wing width, separately for each sex. Sexual difference was examined by unpaired t-test for each character and interspecific difference was examined by Kruskal-Wallis test separately for each sex, except for the four characters representing the body size. Bonferroni correction (Rice, 1989) was applied for coping with probability errors caused by multiple comparisons. For the characters selected by Kruskal-Wallis tests with Bonferroni correction, 10 males and 10 females were measured in each of the remaining six strains. Combining such data of the nine strains, we performed the canonical discriminant analysis, using the computer software STATISTICA (StatSoft, 2000), for the three species, separately for each sex.
The following three species are very similar in general morphology. External qualitative characters commonly seen in all the three species are first referred to in the redescription of D. lini but not repeated in the description of two new species. Quantitative characters that vary among the three species are analyzed in the subsequent section.
Drosophila (Sophophora) lini Bock & Wheeler, 1972
Drosophila (Sophophora) lini Bock & Wheeler, 1972: 59.
Diagnosis. Paramere apically round and with 3-5 minute sensilla, submedially with strong, triangular, inward expansion of which posterior margin only slightly concave (Fig. 2A,C). Aedeagal basal process with fine, irregular ser-rations on slightly wider, apical margin (Fig. 2B).
Description ♂ and ♀. Head: Eyes bright red. Postocellar setae convergent. Supracervical setae tapered, thin, apically slightly curved and sharp. Frons including ocellar triangle pale brown; ocelli orange. Anterior reclinate orbital seta much closer to proclinate seta than to posterior reclinate seta. Antennal pedicel usually yellowish brown. Face and gena yellowish brown. Carina convex, somewhat broader in ♀. Palpus and clypeus brownish yellow. Palpus with 1 prominent terminal and another subprominent, lateromedian setae. Cibarial posterior setae long, tapered, thin, curved and apically sharp; medial and anterior setae shorter than posterior setae.
Thorax: Scutum and thoracic pleura brownish yellow. Acrostichal setulae in 6 rows in front of anterior dorsocentral setae, 4 rows between dorsocentral setae. Basal scutellar setae parallel; apical setae cruciate at right angle.
Legs: Preapical dorsal seta present on tibia of all legs; apical seta on tibia of fore- and midlegs. Longitudinal sex-combs present along entire lengths of 1st and 2nd tarsomeres of ♂ foreleg.
Wing transparent, slightly yellowish. Veins brownish yellow. Basal medial-cubital (bm-cu) crossvein absent.
Abdomen: Second to 6th tergites yellow, each with very distinct apical black band, except for ♂ 6th black dorsally.
Male terminalia: Epandrium pale brown, broad, not pubescent, with triangular expansion covering base of sur-stylus; ventral lobe apically round. Cercus separated from epandrium, not pubescent, triangular; lower, oval part separated, differentiated as secondary clasper. Hypandrium pubescent on caudolateral lobes and caudal margin, caudomedially with strong protrusion apically bearing a pair of stout paramedian setae as long as paramere. Paramere large, longer than wide. Aedeagus slender, apically finely hirsute, basally with a pair of lobate processes as long as aedeagus, fused to apodeme; apodeme rod-like, as long as aedeagus.
Specimens examined. Ten ♂ and 10 ♀ each from the following isofemale strains: BG14028-0581.0, DHS401, DHS501, NKS9212 and NKS9231.
Distribution. China (Taiwan, Guangdong).
Drosophila (Sophophora) ohnishii sp. nov.
Diagnosis. Paramere apically round and with 3-5 minute sensilla, submedially with moderate, inward expansion of which posterior margin distinctly concave (Fig. 2D,F). Aedeagal basal process with fine, irregular serrations on narrower, apical margin than in D. lini (Fig. 2E).
Holotype ♂ from MMY326.
Paratypes. Nine ♂ and 10 ♀ from MMY326; 10 ♂ and 10 ♀ from MMY307.
Distribution. Myanmar (Pyinoolwin).
Relationships. This species is very closely related to D. lini: reciprocal crosses between these two species can produce F1 hybrids, of which females are fertile but males sterile (Oguma et al., 1995). Morphologically, these two species can hardly be distinguished from each other even by the characters of male terminalia which are given in the diagnosis. A combination of some quantitative characters can be used, but not absolutely, to discriminate these two species (see below).
Etymology. Patronym, in honor of Dr. S. Ohnishi who found this species first.
Drosophila (Sophophora) ogumai sp. nov.
Diagnosis. Paramere apically narrow and with 2-3 minute sensilla, submedially with moderate, inward expansion of which posterior margin distinctly concave (Fig. 2G,I). Aedeagal basal process irregular on apical margin but without fine serrations (Fig. 2H).
Holotype ♂ from RGN206.
Paratypes. Nine ♂ and 10 ♀ from RGN206; 10 ♂ and 10 ♀ from RGN3.
Distribution. Myanmar (Yangon).
Relationships. This species is also closely related to the foregoing two species: crosses with them produce F1 fertile females but sterile males, but crosses between ohnishii females and ogumai males produce no F1 hybrids (Oguma et al., 1995). Morphologically, this species can be distinguished from the other two species by the diagnostic characters, but only for males. Combinations of some quantitative characters can be used, not only for males but also for females, to discriminate this species from the others (see below).
Etymology. Patronym, in honor of Dr. Y. Oguma who detected the presence of postmating isolation between this species and the other two species for the first time.
Comparison of quantitative characters
The mean±SD and the range are shown for 42 quantitative characters, separately for each sex and for each species, in Table 3. In interspecific comparison for each sex, there is no character with nonoverlapping ranges between species, which can be used as the specific diagnosis.
Mean±SD and range of 42 quantitative characters in three species
Using the data for the three strains, BG14028-0581.0 (D. lini), MMY326 (D. ohnishii) and RGN206 (D. ogumai), which were measured for all the 42 quantitative characters, correlations with the body size (ThL: thorax length) were examined for 38 characters. Seven characters for male and five characters for female showed significant correlations with ThL in any of the three species (Table 4). However, only FW/HW in male of BG14028-0581.0 held a significant correlation after Bonferroni correction. Since no character was thus regarded as being affected constantly by the body size, i.e. culture conditions, all the characters except for those representing the body size were subjected to the following analyses.
Quantitative characters showing significant correlations with the thorax length in the three sibling species
Significant sexual difference was detected by unpaired t-test for 16 characters in any of the three species (Table 5). Among a total of 27 such cases were seven cases still significant after Bonferroni correction; especially, prorb (relative length of the proclinate orbital seta to the posterior reclinate orbital seta) was significantly different between male and female in all the three species. Therefore, the following analyses were carried out separately for each sex.
Quantitative characters with significant sexual difference (by unpaired t-test) in the three sibling species
Kruskal-Wallis tests detected significant interspecific difference in 23 characters for male and 17 characters for female (Table 6). Of 26 characters that differed significantly among the species in either sex, 15 characters showed still significant interspecific difference after Bonferroni correction. These 15 characters were used in the canonical discriminant analysis.
Quantitative characters with significant difference (by Kruskal-Wallis test) among the three sibling species
Fig. 3 shows plots of canonical scores for measured specimens, resulting from the canonical discriminant analyses, separately for each sex. A total of 90 individuals, 10 each from the nine strains (five strains of D. lini and two each of D. ohnishii and D. ogumai), were measured in each sex, and the 15 characters were used in male and 13 of them, excluding two male-specific characters, in female for the analyses. At a glance of the figure, the first canonical function can be regarded as discriminating D. ogumai from the other species, and the second function as discriminating D. ohnishii from the others, in both sexes. The first function means large ocps (the number of occipital setae) and C3F (the relative length of heavy setation in the third costal section) in both sexes, and large sctlp (the distance between ipsilateral scutellar setae / the cross distance between apical scutellar setae), 4c (the relative length of the third costal section to M1 between r-m and dm-cu), rcorb (the relative length of the anterior reclinate orbital seta to the posterior reclinate orbital seta) and FW/HW (the relative width of frons to head width) in female (Table 7). The second function seems to be reverse in the direction between male and female plots, meaning large FW/HW, small sc2 (the number of teeth in sex-comb of foreleg 2nd tarsomere), 5x (the relative length of CuA1 between dm-cu and wing margin to dmcu between M1 and CuA1), M (the relative length of CuA1 between dm-cu and wing margin to M1 between r-m and dm-cu), C3F and 4v (the relative length of M1 between dmcu and wing margin to M1 between r-m and dm-cu) in male, but large C3F and small FW/HW in female (Table 7). Classification functions for the three species, based on equal a priori probability for each species, are shown in Table 7, and the results of posterior classifications of the 90 cases (specimens) in each sex are shown in Table 8. The percentage of correct classifications was rather high in both sexes, 86.7% in male and 91.1% in female. Thus, the present results suggest that the 15 (for male) or 13 (for female) quantitative characters in combination are effective to discriminate the three sibling species, even their females, from each other with considerable confidence.
Factor structure coefficients (correlations between the characters and the discriminant functions) and classification functions for the three sibling species, resulting from canonical discriminant analysis in which classifications were based on equal a priori probability for each species
Summary of posterior classifications. Rows: observed classifications, columns: predicted classifications
In this study, we examined morphological differences among three sibling species that had proved to be biologically good species reproductively isolated from each other almost completely (Kim et al., 1989; Oguma et al., 1995). Of those three species, two were described as new species, i.e. D. ohnishii and D. ogumai, and the remaining known species, D. lini, was redescribed in the light of detailed examination of the male terminalia and many quantitative characters. However, they are very similar in morphology, hardly distinguishable from each other even by a few characters designated as the diagnosis, especially between D. lini and D. ohnishii, implying that they are still in the process of species differentiation, even though having reached almost to the level of good species. This is supported also by partial crossability and fertility of F1 hybrids between them (Kim et al., 1989; Oguma et al., 1995).
It is conceived, but implicitly, that character differentiation does not always proceed in parallel among different biological properties in the process of speciation. Here, morphological differentiation among the three sibling species is compared with the degrees of premating and postmating isolation among them, based on the data presented by Oguma et al. (1995) for the reproductive isolation. Morphological difference was evaluated by the squared Mahalanobis distance based on 15 characters used in the canonical discriminant analysis for male and 13 such characters for female. Oguma et al. (1995) provided quantitative data for premating isolation based on their pair-mating experiment. They set up 10 replicates (i.e. pairs) for each cross and observed mating behavior of each pair until copulation occurred within five minutes. We converted their data of copulation frequency (the number of pairs having copulated within five min., Table 2 in Oguma et al., 1995) into those of copulation rate per pair (Table 9A), and calculated the following premating isolation index between different species:
Degrees of premating and postmating isolation based on the data of Oguma et al. (1995)Oguma et al. (1995) presented the data in qualitative categories representing different degrees of isolation (Fig. 4 in Oguma et al., 1995). We converted such categories into relative values: “F” (female and male F1 hybrids fertile)=0, “f” (female F1 hybrids fertile but male F1 hybrids sterile)=0.5, and “(–)” (no F1 hybrids produced) = 1 (Table 9B). Such values of all reciprocal crosses between different species were averaged for each pair of species. Relationships among the three species were represented by a triangle of which side lengths corresponded to the interspecific differences or degrees of isolation (Fig. 4).
At a glance of the figure, we notice that the interspecific differentiation patterns vary among the concerned biological properties. In morphology, D. ogumai is most remote from the other two species for both sexes. On the other hand, premating isolation is almost complete between D. ohnishii and the other two species, but moderate between D. ogumai and D. lini. Oguma et al. (1995) reported that females of D. ohnishii performed strong repelling behaviors, spreading and fluttering their wings or kicking males, against allospecific males and D. ohnishii males were strongly refused by allospecific females. Postmating isolation is more or less present between the three sibling species, heterospecific crosses usually producing F1 fertile females but sterile males. However, some crosses between D. ohnishii females and D. ogumai males produce no F1 hybrids. The pattern for premating isolation suggests two possible hypotheses. One possibility is that this property has been evolved or reinforced in sympatric populations between D. ohnishii and D. lini and between D. ohnishii and D. ogumai. Although no such sympatric populations have been found yet, it is probable that the range of D. ohnishii overlaps with that of D. lini in southwestern China to northern Myanmar and with that of D. ogumai somewhere between Pyinoolwin and Yangon, which are not so distant from each other. The other possibility is that the strong premating isolation between D. ohnishii and the other two species has evolved in a very restricted population at a locality of central Myanmar, perhaps by a few mutations.
Thus, the present study provides very interesting materials and information for studies of speciation mechanisms. To fully understand the evolution of the three sibling species, however, much more information is needed, especially about some gene sequences, premating isolation (i.e. mate discrimination) mechanisms, geographic distribution ranges and eco-physiological adaptations to environmental conditions in their ranges.
We thank Dr. T. Aotsuka (Tokyo Metropolitan University) for providing us with the specimens from the study culture strains and Dr. A. J. Davis (Hokkaido University) for his help in linguistic improvement of the manuscript. This work was supported in part by a Grant-in-Aid for Scientific Research from Japan Society for the Promotion of Science (No. 12375002).